Museum video captions. From your Mac to the gallery.

Start with the exhibit player

A gallery visitor may arrive halfway through a video or watch it without hearing the soundtrack. Before creating captions, ask the installation team how text will appear: a separate track the player displays, or captions permanently rendered into the video. Confirm the accepted format and how captions remain available when there is no remote or caption button.

Polyglot can create a speech-derived subtitle draft locally on an Apple Silicon Mac and export SRT. It does not configure exhibit players or burn captions into video. If the installation needs a rendered copy, include a separate finishing step in your video editor. The National Park Service’s audiovisual guidance distinguishes captions, audio description and assistive listening; a subtitle file alone is not a complete exhibit accessibility plan.

Pair the final edit with the right permissions

Use a recording you own or have permission to process for this exhibit. Confirm that the intended display, translation and any later web reuse fit the project’s permissions, including those for speakers and included artwork or music. Access to an archive file is not a reason to assume unrestricted reuse. Keep those decisions in the exhibition’s rights record, outside the subtitle text.

Work from the exact final clip, including its opening title and closing pause. Keep the unchanged source and give the edit a stable identifier such as gallery-loom-edit03. Prepare a short reference list of object names, makers, place names and dates from approved collection records. The list helps a human reviewer; it is not a claim that Polyglot imports a custom museum vocabulary.

Review the object story with and without sound

Import the narrated clip, create the original-language draft and compare its text with playback. Speech recognition supports 25 European languages; translation targets are a separate capability. Local media and text processing still involves model setup and separate network boundaries described in our privacy information. A silent clip with no speech needs a different authoring workflow, not speech recognition.

For a fictional loom demonstration, check whether the narrator says “warp” or “weft” and “1880” or “1890.” Preserve what was actually said; if it conflicts with the approved label, flag the recording for curatorial review instead of silently rewriting a quotation. A bell heard off screen may also explain why the demonstrator stops. W3C caption guidance includes meaningful sounds and needed speaker context, not only dialogue.

Polyglot does not automatically identify speakers or label sound events. Have a caption reviewer decide what context the audience needs. Add new sound-only cues, adjust timing or reposition captions in a compatible finishing editor where required. Watch again with sound off: the visitor should be able to understand why the action changes, not merely read the narrator’s words.

Translate the meaning and check the reading time

Review each additional language with someone who understands the exhibit’s subject. Agree how collection titles, historical names and specialist terms should appear. A compact object label may use a different phrasing from spoken narration; preserve the distinction where it matters. Keep an approval note for each language rather than assuming that an approved source track approves every translation.

Polyglot preserves source cue timing in translated text. Read longer translated lines while the clip plays, especially before a close-up, title or pause. If the text cannot be read comfortably, revise it with the reviewer or finish the timing and segmentation externally. Do not shorten dates, qualifications or culturally significant names just to fit an arbitrary line length.

Hand off the file the installation can use

Export reviewed SRT files and label each with the matching edit and language, such as gallery-loom-edit03.en.srt. Ask the installer to confirm support before relying on a sidecar file. For a browser-based display using HTML text tracks, MDN documents WebVTT as the track format. Convert outside Polyglot when needed and compare cue count, order, text and timestamps with the reviewed export; changing the extension is not conversion.

For always-visible captions rendered into the picture, retain both the caption-free master and the finished display version. Set appearance in the finishing workflow and inspect it on the actual screen. Polyglot’s SRT export does not guarantee font size, contrast, placement, player behavior or support for multiple language tracks on a particular kiosk.

Accept the display, including its restart

Run the matching media and captions on the installed display for at least two complete loops. Check the first and last cue, the closing pause and the return to the beginning. Text from the ending must not linger over the next opening. Restart the player using the installation’s normal procedure and confirm that the intended caption track still appears without staff intervention.

Read from the intended viewing positions under gallery lighting, with sound off as well as on. Check that text does not obscure object details or on-screen labels, and inspect accents and punctuation in every delivered language. These are suggested acceptance checks, not a claim that Polyglot has been tested with your hardware. Record the media edit, subtitle revision, player configuration, reviewer and any remaining issues in the handoff. A replacement video or player update needs another display check.

Keep exploring.